Want to know a secret?

‍

I'll show you how I used ChatGPT to write yesterday’s post.

‍

Not entirely, but enough to save hours of research and polish.

‍

Yesterday, I wrote about how social media is shifting towards paid subscriptions. To back up my argument, I needed a full list of subscription products from major platforms.

‍

Rather than spending hours collecting data, I used Deep Research on ChatGPT to generate a comprehensive table with all the details.

‍

I used a short and simple prompt, no complex prompt engineering here.

‍

I quickly drafted my post, then I ran it through LinkedIn GPT.

LinkedIn GPT is a custom GPT I built by uploading all my past LinkedIn posts (yes, all of them!) plus some content from my favorite creators.

It outputs a refined, high-impact post that still sounds like me.

‍

Many creators want you to buy complex templates and courses on how to automate your life with AI. But this is not the point.

‍

AI doesn’t have to be complicated. If it does feel complex, you’re probably doing it wrong.

It should be so simple that it becomes part of your daily workflow, as a personal assistant rather than a replacement (for now).

‍

How do you use AI in your daily work? 👇

ChatGPT Deep Research is now available to all Plus users!

‍

Two weeks after I got my Pro $200/month subscription 😭

‍

Deep Research was by far the most attractive feature of GPT Pro.

I used it a lot and saved countless hours already.

‍

Soon after its launch, it was copied by Perplexity and included in its $20/month subscription.

‍

Then it was copied by xAI Grok, while Google already had its own version in Gemini Advanced.

‍

GPT Plus gives 10 Deep Research queries per month, which is not a lot, but already enough for most users.

GPT Pro gives 120 queries per month, more than you'll ever need I think.

‍

Now it'll be all about who provides the best Deep Research results!

‍

So far, I have tried GPT Pro and Perplexity and the former still wins.

‍

But we'll see!

💰The internet is going premium.

‍

We're witnessing a fundamental business shift:

‍

The ad-supported internet is giving way to a subscription-based model.

‍

Just look at social media:

‍

‍LinkedIn Premium: $29.99+ / month

‍X Premium: $3+ / month

‍Meta Verified: $11.99+ / month

‍YouTube Premium: $13.99 / month

‍Snap Inc.+: $3.99 / month

‍Telegram Messenger Premium: $4.99 / month

‍Reddit, Inc. Premium: $5.99 / month

‍

Even TikTok is testing an ad-free $4.99/month subscription!

‍

Search is no different:

‍

‍Perplexity Pro: $20 / month (soon launching its own premium? browser)

‍ChatGPT DeepSearch: $200 / month

‍

Soon, paying for a "good enough" version of the internet will be the norm.

‍

This changes everything:

‍

Digital marketing, advertising, website UX, etc

‍

Are we ready for a world where the best online experience comes with a price tag?

Perplexity is launching its own browser!

‍

It makes a lot of sense. Here’s why: 👇

‍

Yesterday, Perplexity announced its new “Agentic” browser: Comet.

‍

The strategy is to own the entire user journey, from search to browsing, and even to shopping with features like "Shop like a Pro."

‍

Pro subscriptions and shopping can be profitable, but last November Perplexity admitted that "subscriptions alone do not generate enough revenue."

‍

So, where do you find extra revenue streams?

‍

Advertising, of course!

‍

Perplexity has already started to test sponsored follow-up questions in its AI chatbot. But with its own browser, it’s going all in. 🚀

‍

Owning a browser means owning:

‍

↳ How cookies and user data are stored.

↳ How ads are targeted, displayed, and monetised

↳ How users interact with search and shopping.

‍

Even social media giants like Meta and TikTok route external links through their own in-app browsers.

‍

AI is supposed to change everything…

‍

Yet, AI companies are falling back on traditional business models, like advertising.

‍

The big question is:

‍

Why hasn’t Google done this with Chrome and Gemini? 🤔

‍

Not yet, but they will. Inevitably.

‍

At that point, will Perplexity still stand a chance?

‍

AI booked me a flight while I was… 🚽💩 busy.

‍

That’s the real power of AI agents. Not speed, but delegation. They free us up to do anything else. Yes, even that 😅.

‍

Take Operator by OpenAI.

‍

I gave it a simple instruction:

‍

"Find me the cheapest direct flight from X to Y."

‍

It didn’t just fetch a list of flights. It reasoned, researched, and optimised, scanning multiple airports to find the best option.

‍

Next step would be including train fares to and from the airport for a real total cost comparison.

‍

Unlike automations, AI agents interact in real time:

‍

→ They ask for clarifications.

→ They offer multiple options.

→ You can change scope on the fly, like when I switched flight dates mid-process.

‍

My booking didn’t actually go through because Operator doesn’t store my personal or payment details, so it asked for me to take over.

‍

But I'm sure there's a way to fix this.

‍

A few interesting facts for marketers:

‍

-> Operator browses from the US.

Not sure if it's because I use a VPN to access it, or it does it regardless. But it's likely agents will be primarily US-focused. This is a potential challenge for European marketers, who will need to optimise for US as well.

‍

-> It relies on Bing for search.

Your SEO strategy is better be strong on Bing too, not just Google.

‍

AI agents are still in their early days, but it's now clear where we’re headed.

‍

Marketers, are you paying attention?

Zapier automations are NOT AI agents.

‍

And yet, many people sell automations as "AI agents." They’re not.

‍

Here’s the difference:

‍

What an AI agent actually is:

‍

A system that, given a problem (even a vague one), figures out the best way to solve it. It decides its own steps and executes them autonomously.

‍

Here's how an AI agent and a regular automation solve the same task:

‍

“Book the cheapest flight from Switzerland to London on May 7th.”

‍

👉 AI agent:

It searches the web, compares multiple flights dynamically, finds the best option, and (ideally) books it.

‍

👉 Zapier automation:

A rigid, predefined workflow that follows explicit step-by-step instructions.

‍

Steps:

1️⃣ Get ChatGPT to generate Python code to scrape Skyscanner.

2️⃣ Run the code (fingers crossed it works).

3️⃣ Scrape Zurich-Heathrow flights.

4️⃣ Scrape Zurich-Stansted flights.

5️⃣ Scrape Geneva flights…

🔹 … Step 15: Get ChatGPT to determine the cheapest flight.

🔹 … Step 16: Open Skyscanner and book it yourself.

‍

Key difference:

‍

Automations follow preset instructions, you tell them exactly what to do. AI agents figure things out on their own.

‍

Both can solve the problem. But agents are flexible, like a human.

‍

For example, if you want to change the flight date with an AI agent, you simply update your initial prompt. The same as changing the dates on Skyscanner.

‍

With Zapier, you need to edit the entire workflow.

‍

Have you tried some of these?

Elon never wanted to buy OpenAI.

‍

Instead, this is what his $97.4 billion offer really meant:

‍

🔹 OpenAI is transitioning from non-profit to for-profit.

To do so, a new for-profit entity needs to acquire OpenAI’s assets from the non-profit, ideally at the lowest possible price.

Since OpenAI is still burning billions, cash is a concern. And when you're a private company, you can value your assets however you like (kind of).

$1 inter-company transactions are not unheard of.

‍

🔹 But Elon is messing with Sam's plans.

His formal $97.4B offer put a price tag on OpenAI’s assets.

Now, it’s much harder for Sam Altman to justify a low valuation for the transition.

Higher price means more cash to raise and slower progress.

‍

🔹 Elon's team also committed to keeping OpenAI non-profit,

right when Sam is pivoting it to a for-profit entity.

Not only that, they will even withdraw the offer should OpenAI decided to stop the transition.

The message is clear: Elon is the "good guy", prioritising humanity over profits, unlike Altman.

‍

🔹 Then, the killer move: due diligence.

The offer letter included a May 10th deadline and demanded access to OpenAI’s "assets, facilities, equipment, books, and records" as well as "particular personnel involved in the Business" for interviews.

Even if the deal didn't go through, meanwhile xAI would get invaluable insights into its competitor’s business and technology.

‍

🔹Elon got exactly what he wanted.

Few days ago, OpenAI’s board formally rejected the offer.

But Elon still got what he wanted:

‍

- Undermine Sam Altman's reputation.

- Regulatory scrutiny over the for-profit transition.

- Slowdown in OpenAI operations and plans.

‍

Often dismissed as just erratic and crazy, Elon is actually playing a long and calculated game. Later today he will release the latest xAI models, supposed to be the best of all.

‍

I wouldn't be surprised if Elon eventually won the AI race, same as he did with EVs and space.

‍

Time will tell!

The US is not at the forefront of EVs.

‍

Will this give an edge to Europe at some point?

Forget SEO.

‍

Welcome to AEO (Agent Engine Optimisation).

‍

I tested OpenAI Operator by asking to find a suitable apartment for rent in Zurich within a given budget.

‍

Operator opened its browser.

‍

It used Google Chrome, but it didn’t search on Google.

‍

It searched on Bing (no surprise here, given Microsoft’s massive investment).

‍

Then things got interesting.Instead of heading to Homegate or ImmoScout24 (specialised real estate platforms in Switzerland, owned by my employer), it went straight to comparis.ch AG, a general comparison site and my employer's competitor.

‍

I noticed a similar pattern when searching for flights.

‍

Rather than Skyscanner, it often defaults to Google Flights.

‍

How can you make sure agents choose your website over competitors'?

‍

Does it even matter?

‍

Well, it does.

‍

At the end of the process, which lasted about 10 minutes, Operator returned two links of two suitable apartments. I clicked on them and eventually interacted with the website myself.

‍

💡You want to make sure those links bring to your website!

‍

In a world where agents browse on behalf of humans, SEO and marketing strategies will need to evolve.

‍

UI, UX, copywriting and SEO will have to shift from human-centric to agent-centric design. Remember, agents don't feel emotions! This changes everything.

‍

Are marketers ready for this?

I went all in! 🔥

‍

I subscribed to ChatGPT-Pro for $200/month.

‍

Like someone way smarter than me once said:

"The risk of under-investing is dramatically greater than the risk of over-investing."

— Sundar Pichai, CEO of Alphabet Inc.

‍

And I believe this is exactly the case with #AI.

‍

I've been experimenting with AI tools for over two years, but only now do I feel something fundamentally shifting.

‍

The o1 model, released last autumn, is years ahead of GPT-4o.

It's so good that I kept hitting the usage limit with my Plus subscription.

‍

So, I subscribed to Claude to try their latest 3.5 Sonnet model.

Same issue, kept hitting the limit.

‍

Tried DeepSeek... good quality, when it works.

It’s often unavailable and lacks the polished UX of competitors. No custom GPTs, no Claude Projects etc.

‍

Then, I discovered that GPT-Pro unlocks an even more advanced model: o1-pro, with no usage limits. 🤯

As if that wasn’t enough, OpenAI dropped Deep Search last week, meant to be a complete game changer.

Add Sora and Operator (via VPN), and I was sold.

‍

First impressions: blown away.

‍

Deep Search + o1-pro is the real deal.

In minutes, it completed research tasks I had worked on for months.

‍

Sora seems far from the mind-blowing demos we saw months ago.

I’d say it’s comparable with Runway. But, for $200/month it's virtually unlimited.

Try replicating the same projects with Runway, and you'll easily spend the same amount or more. A single video project I worked on in September costed me around $70. So, all in all Sora is a good deal.

‍

Operator is also a bit underwhelming, but gives a clear idea of where Agents are heading.

‍

Not sure how long I’ll keep the subscription, but for now I’m all in.

‍

Over the next few weeks, I’ll be posting some real-life use cases of GPT-Pro in the context of content and marketing.

‍

Stay tuned!